Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Proba-V QWG-07
Proba-V and S3-SYN SNAP Toolbox: status and updates
Carsten Brockmann
04.05.2018
SNAP 6 Release January 2018
SNAP Version 6 http://step.esa.int
• Direct data access (SciHub) integration • Plotting of metadata values • Ocean Colour (C2RCC) processor • Radiometry tool (Rad2Refl) for SLSTR
Proba-V Toolbox
• Publicly available through ESA STEP Website
• Distribution with SNAP5, December 2016 • Version 2.0 with SNAP6, Jan 2018
• Proba-V reader for L2A and L3 synthesis products
• RGB support (PB-V profiles)
• Enabling all image visualisation, analysis and processing function • Non exhaustive list: band math, projection, collocation, mosaicking, statistics, extraction, filtering,
subsetting, binning, resampling, classification, segmentation, format conversion
Proba-V in STEP Forum
iCOR @ SNAP
• iCOR for Sentinel 2 and Landsat-8
• VITO development, integrated into SNAP via the Stand-Alone Adapter
• Distribution through VITO Website
• Integration into SNAP Desktop
• Support through STEP Website & VITO
iCOR in SNAP Forum
Prioritised features for PBV-TBX Evolution
• Follow-on from last QWG • QWG06: presentation of SNAP processors (S3, S2) which have the potential to
be applied to Proba-V • Subsequent analysis and feedback by VITO
• Idepix Cloud Masking: Give the user the option to experiment with different cloud masking algorithm suitable for his or her needs.
• Soil & Vegetation Radiometric & Water Indices: Quite some indices are available via the Copernicus Global Land Service, so no need to invent the wheel again. Only the indices that are not available are a nice plus for PROBA-V users
SNAP Evolution 2018 – 2020
SNAP http://step.esa.int
4 major releases in next 2 years
• Time series exploration • Cloud access – data and processing • Improved SNAPPY • OLCI Smile Correction • OLCI Atmospheric Correction • Water Quality Operators • OLCI & SLSTR Synergy L1C Tool
Cloud Support
• SNAP Processing Services including SNAP Engine Server backend (WPS)
• SNAP Desktop GUI frontend.
• Demo servers with data local WPS, close to Sentinel data.
• IPython-like Remote SNAP REPL (Java 9) interface
Improve integration of Python
• Support of individual Python environments
• SNAP Desktop: Python Plugin Manager to add plugin paths and to configure python interpreter and to create distributable Python plugin bundles
• Add “pythonic” API to better support the python developers and ease their life
• Allow Python function in Band Math • SNAP’s band math expression editor in this case becomes a Python code editor. • Changes in the Python code shall be reflected immediately in a target band’s pixel
data. • Displayed image of this band will be immediately updated (“hot deployment”).
• Python REPL as a command-line interface window in SNAP Desktop. • Ideally, this integration would be based on the Jupyter / IPython notebook
New Standard IO Format
• Lazy loading
• Reuse the source binaries
• User selectable data format
• Multi-size images, image tiling, image pyramids
• Single file ZIP
• Support different user requirements
• Support of cloud storage
• Store operation instead of data
• Interoperability with GIS software
Enhance GPF capabilities
• Support any workflow step types • Operator graph (current state)
• Operating system commands
• invoking a remote web processing service
• Support various output types • New product instances (current state)
• Text files
• Vector data
• Plots / images / movies
Further exploit multi-size product model
• Support multi-size products in GPF operator API.
• Add multi-resolution data support to operators: Subset, Binning, Mosaicking, Reprojection, Colocation
• Write and read multi-size products to/from other formats than DIMAP
• Add multi-resolution data support to SNAP Desktop functions: “Copy Pixel Info to Clip Board”, “Create Subset”, “Export Mask Pixels”, “Magic Wand Tool”, “Export Transect Pixels”, “Transfer Mask”
Introduce global resources library
Including access to online geospatial data: OPeNDAP, WCS, WCPS
Time series support
• Virtual stacks of products • Product groups with a certain attribute that orders them (e.g. time)
• Time series tools • Time series operator – prerequisite is common spatial grid
• Visualisation of variables along time axis
• Time series matrix
• Scrolling through time in image stacks
• Managing time series (add, edit, remove products from time series)
Improved support for uncertainties
• Promoting existing functions • Error propagation in band maths
• Visualisation
• tutorials, training
• Monte-Carlo Propagator • Associate uncertainties to input of an operator
• Ensemble generation
• Support sensitivity studies
OLCI/SLSTR Level-1C SYN Tool
L1C-SYN Tool for SNAP
• Users need Synergy product on Level 1b in order to perform individual Level-2 processing • E.g. Copernicus Global Land Service and ESA CCI projects
• Requirements shall be fulfilled by a user tool. • A tool gives more freedom to the users compared
to pre-defined product from the IPF • Define and configure own L1C-SYN product
OLCI / SLSTR Level-2 Synergy (will) exists - Why a L1C-SYN Tool?
L1C-SYN Tool - Stages
1 2
3
• Combination of existing operations
• Selection of bands • Spatial subsetting &
resampling • Release: Soon
• MISR file used for improved co-registration
• Export to a tile grid Sentinel-2/Proba-V
• Release: Summer 2018
• Different co-registration methods
• Usable with other Sensors • Release: End 2018
L1C-SYN Tool - Processor
Sentinel 3 Land Product Level 3 fAPAR (OGVI)
OLCI Terrestrial Chlorophyll Index (OTCI)
Intercomparison of S3 and Proba-V Land Products
• Sentinel 3 Level 3 products of OTCI and OGVI (fAPAR) for validation and evaluation • Activity started within S3 MPC to support Land Validation
Activities • Protoype presented at S3VT (March 2018) • Continuation under discussion with ESA
• Product Definition
• MODIS Level 3 grid to allow easy comparison,
analysis and further usage of products from both
sensors.
• Global coverage at spatial resolution of 500m
• Sinusoidal projection
• 8-day and monthly temporal aggregation periods
• Tiling on MODIS tile grid
bs fAPAR (OGVI) monthly mean April 2017
Consistency Test with MERIS
MTCI
(0 to 6)
OTCI
(0 to 6)
August
Septem
ber
October
Difference
(-2 to +2)
Work performed by J. Dash/L. Brown USouthampton
Proba-V Symposium
• Demonstration during lunch time on Tuesday and Wednesday
• Interactive presentation
• Questions & Answers
• Topics: • Validation with SNAP
• Generic operations on data
• SNAP & Python
• Cloud Screening with IdePix and what it can offer to Proba-V
• Synergy between Proba-V and S3 OLCI / S2 MSI
Standard Neural Net Format
Standard Neural Net Format (NNF)
• Description of the architecture of neural nets • Input / output neurons
• Hidden layers
• Neurons
• Activation functions
• Weights
• Library for reading and executing neural nets • Neural nets become auxiliary data (ADF)
• Software processor is independet from actual net
• Independecy • Code and ADF kept separate allowing faster updates
• No technical dependency from a certain net provider
Support by NNF
• Neural Net Types • Multilayer Perceptrons (MLP) • Radial Basis Function nets (RBF)
• Activation Functions • Sigmoid • …
• Additional layers for pre- and postprocessing • Input/output scaling • Normalisation • Linear combinations • …
• Library • C++ version (used in MERIS and OLCI ground segment) • Java version (used in BEAM and SNAP)
Example
Summary - Recommendations
1. Evolution of Proba-V Toolbox • Add Proba-V support to IdePix
• Add Proba-V support to those Soil & Vegetation Radiometric & Water Indices which are not supported by Global Land Service
2. Add Proba-V Grid to Sentinel 3 Land Level 3 production
3. Use NNF for Proba-V • Luis to write his net in NNF format
• Vito to use NNF library in processor
• Prerequisite: BC to update the format in order to stay up-to-date with recent developments (CNN, Tensorflow, …)